Files
Chica_08762300_2025.pdf
Open access - Adobe PDF
- 24.51 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- This thesis builds and validates an end-to-end pipeline for rapid, repeatable evaluation of turbine blade aerodynamics using steady RANS in SU2 v8.0.1 (“Harrier”). The workflow, from mesh generation to solver execution and automated post-processing, is exercised first on the SPLEEN C1 cascade and then on a set of machine-generated blades used as a screening dataset. Two turbulence closures are considered throughout, SA-BCM and κ-ω-SST-LM. Methodologically, the meshes target y+ ≈ 1 with a mild growth ratio (r ∈[1.1, 1.2]) and BL thickness estimated from δ99, with wake sampling aligned to the experimental plane (PL06). Validation metrics include the RMS of the isentropic surface Mach, wake total-pressure loss at PL06, and surface skin friction Cf . Mesh effects are assessed via grid convergence index (GCI) alongside a small-change criterion; computational cost is reported in Core-hours to enable fair comparisons. On SPLEEN, the pipeline reproduces the main experimental trends across operating points and provides clear guidance on mesh knobs and model choice. For batch screening, SA-BCM offers robust convergence and faithful blade load and wake behavior at modest cost. The latter κ-ω-SST-LM model is retained as a second option when wake metrics under demanding regimes are the priority. Applied to the blade dataset, the pipeline exposes failure modes early (TE/wake discrepancies, divergence) and supports data-driven acceptance/rejection with minimal user intervention. The study is intentionally pragmatic: results are 2D with clean inlet conditions, and the focus is on a conservative, transferable setup that balances accuracy and throughput. Within this scope, the pipeline closes the loop between fast blade generation and physics-based evaluation, and constitutes a practical screening layer that can be extended to 3D endwall effects and transition modeling in future work.